• Tech Support ⤴
  • Projects
  • Services
    • AI Development
    • UI/UX Design
    • Web Development
    • Technology Support
    • Mobile App Development
    • Banking ATM Interfaces
    • Process Automation
    • Security Auditing
    • Local AI Servers
  • odoo ERP
get in touchStart with Eva
logo
Tech Support ⤴
Projects
Services
AI DevelopmentUI/UX DesignWeb DevelopmentTechnology SupportMobile App DevelopmentBanking ATM InterfacesProcess AutomationSecurity AuditingLocal AI Servers
odoo ERP
get in touchStart with Eva
Loading…
logo

Transforming businesses through AI-powered digital innovation and creative excellence.

Quick Links

BlogAinexProjectsContact us

Contact Us

pinDubai Digital Park, A5, DTEC - Silicon Oasisemail[email protected]phone+971 55 7538087
© 2026 aratech. All rights reserved.
Privacy PolicyTerms of ServiceCookie Policy
Home / Blog / HyperFrames: The Open-Source Framework That Turns HTML Into Production Video

HyperFrames: The Open-Source Framework That Turns HTML Into Production Video

HyperFrames is a fresh take on programmatic video: write plain HTML, style with CSS, and render frame-accurate MP4s — purpose-built for AI agents and

May 20, 2026 - 11 min read

Key Takeaways

ExpandCollapse
  • - HyperFrames turns HTML and CSS into frame-accurate MP4 video programmatically
  • - Designed for AI agents and developer pipelines that need video at scale
  • - Open-source framework with a familiar web stack instead of proprietary video APIs
  • - Supports composition, styling, and rendering in automated content workflows
  • - Fits teams building dynamic video for product, marketing, or agent outputs
HyperFrames HTML composition code side-by-side with rendered MP4 output, showing HTML to video workflow

Table of Contents

  • Introduction
  • The Problem With Programmatic Video Today
  • The Solution: HyperFrames
  • Key Features
    • HTML-Native Composition
    • AI-First Design
    • 50+ Ready-to-Use Blocks
    • The Frame Adapter Pattern
    • Deterministic Rendering Engine
    • Fully Open Source (Apache 2.0)
  • Getting Started
    • Prerequisites
    • Quick Setup
    • First Composition
  • Advanced Usage
    • GSAP Timeline Animations
    • Multi-Track Compositions
    • Website-to-Video Pipeline
    • Lottie and Three.js
  • Comparison: HyperFrames vs Remotion
  • Why This Matters Now
  • Conclusion & Next Steps

Introduction

!HyperFrames rendering pipeline: HTML template -> frame extraction -> video assembly

If you've ever tried to automate video production, you've hit the same wall: every tool — Premiere, After Effects, even web-based compositors — is built for a person sitting at a desk with a mouse, not for a script that runs unattended in a CI pipeline.

What if generating videos worked the same way building a website does?

Today we're looking at HyperFrames — HeyGen's open-source rendering framework that treats HTML/CSS as a first-class video authoring format. You write standard markup, add a few data attributes for timing, and render it into a frame-accurate MP4. No React DSL. No proprietary scene graph. No video editor.

This is the tool that lets AI agents like Claude Code write and iterate on video compositions the same way they write code. And it's Apache 2.0-licensed, fully open source, at github.com/heygen-com/hyperframes.


The Problem With Programmatic Video Today

Video has resisted the "code-as-infrastructure" treatment for decades. Every attempt requires either:

  • A GUI editor that can't be scripted and requires manual export
  • A React wrapper (looking at you, Remotion) that adds a build step and opinionated DSL
  • Screen recording of a browser window — brittle, low quality, frame drops
  • Cloud-render services that cost per minute and lock you into a vendor

The real cost isn't just tooling. It's that video workflows can't participate in Git. You can't diff a Premiere project. You can't review a video composition in a pull request. You can't reproduce the same output from the same inputs unless your artist happens to do it exactly the same way every time. And AI agents — which have gotten remarkably good at writing code — are completely shut out of the medium.

Automated video at scale requires three things that no existing tool fully delivers:

  1. Deterministic output — same input = identical video, always
  2. Git-trackable source — code diffs, rollbacks, branching
  3. Agent-native authoring — LLMs can write and modify compositions without specialized training

HyperFrames is built to deliver all three.


The Solution: HyperFrames

HyperFrames is an open-source video rendering framework from HeyGen that lets you define video compositions as HTML files and render them to MP4 via headless Chrome + FFmpeg.

Its core philosophy is elegantly simple: "Write HTML. Render video. Built for agents."

Every element in your HTML becomes a timed "clip" on the video timeline, controlled through data- attributes. Animations live in standard CSS/JS (GSAP, Lottie, Three.js). Preview is live-reload browser. Render is FFmpeg-encoded, frame-accurate output. No proprietary format stands between you and the final file.

At 19.9k stars and 1.9k forks with 145 releases as of May 2026, it's already the most momentum-rich open-source video framework shipping in the developer + AI agent space.


Key Features

HTML-Native Composition

HyperFrames compositions are plain HTML files. No React. No TSX. No bundler.

<div id="stage"
     data-composition-id="my-video"
     data-start="0"
     data-width="1920"
     data-height="1080">

  <video
    id="clip-1"
    data-start="0"
    data-duration="5"
    data-track-index="0"
    src="intro.mp4"
    muted
    playsinline
  ></video>

  <img
    id="overlay"
    class="clip"
    data-start="2"
    data-duration="3"
    data-track-index="1"
    src="logo.png"
  />

  <audio
    id="bg-music"
    data-start="0"
    data-duration="9"
    data-track-index="2"
    data-volume="0.5"
    src="music.wav"
  ></audio>
</div>

The Rule of Three governs valid compositions:

  • Root element needs data-composition-id, data-width, data-height
  • Timed elements need class="clip", data-start, data-duration, data-track-index
  • Animations must use { paused: true } on GSAP timelines and register on window.__timelines

This is near-zero ceremony. Any web developer who can write div elements can write a video composition.


AI-First Design

HyperFrames isn't just AI-compatible — it's built for AI coding agents.

The CLI is non-interactive by default, which means it's scriptable by design. Error messages are machine-readable. Output is plain text. There are no TTY prompts that block a headless agent.

More importantly, the framework ships with a skills package that teaches AI agents the correct patterns:

npx skills add heygen-com/hyperframes

Once installed, the following slash commands become available in Claude Code:

  • /hyperframes — author and edit compositions
  • /hyperframes-cli — init, lint, preview, render
  • /hyperframes-media — TTS, transcription, background removal
  • /gsap, /lottie, /three, /animejs, /css-animations, /waapi — adapter-specific animation skills

A cold-start prompt like "create a 10-second product intro with a fade-in title and background music" produces a working composition. Warm-start prompts that take existing HTML/PDF/CSV and turn them into videos work equally well. Iteration is just natural language: "make the title 2x bigger, add a fade-out at the end."

This is a qualitatively different experience from trying to drive Premiere via API or hand-coding JSX in Remotion.


50+ Ready-to-Use Blocks

HyperFrames ships an extensible block registry. Install common components with a single CLI command:

npx hyperframes add flash-through-white   # shader transition
npx hyperframes add instagram-follow      # social overlay
npx hyperframes add data-chart            # animated chart

Templates for video styles include swiss-grid, nyt-graph, play-mode, and vignelli — covering corporate clean, editorial data storytelling, energetic social media, and 9:16 portrait formats.


The Frame Adapter Pattern

HyperFrames introduces a "Frame Adapter" abstraction that cleanly separates timing logic from animation runtime. Bring your own seekable animation library — GSAP, Lottie, CSS keyframes, Three.js, or Web Animations API — and HyperFrames handles the frame seeking and compositing.

This means you don't have to rewrite your existing animation knowledge. If your team already uses GSAP timelines for web animations, those same timelines work in HyperFrames video compositions.


Deterministic Rendering Engine

The rendering pipeline is frame-accurate by construction:

HTML Source → Headless Chrome loads composition → Chrome seeks to frame = floor(time × FPS) → screen capture → FFmpeg encode → MP4

Chrome's beginFrame API gives sub-frame precision. No wall-clock playback. No race conditions between playback head and capture. Same HTML, same inputs, same video file every time.

This determinism is what makes AI-agent-driven workflows viable. An LLM can safely iteratively modify a composition, re-render, and do visual diffing — something that's impossible with tools whose output changes based on machine load, clock drift, or the time of day.


Fully Open Source (Apache 2.0)

HeyGen ships HyperFrames under Apache 2.0 — a permissive, fully OSI-approved license. No paid tier above a small-team threshold (as Remotion uses). No source-available licensing. Full rights to fork, modify, ship commercially, or build a product on top of it.


Getting Started

Prerequisites

node -v    # needs >= 22 (check with nvm)
ffmpeg -version  # needed for encoding

Quick Setup

# Option 1: interactive project init
npx hyperframes init my-video
cd my-video
npx hyperframes preview   # opens browser with live reload
npx hyperframes render     # outputs MP4

## Option 2: AI agent (recommended)
npx skills add heygen-com/hyperframes
## Then use /hyperframes in Claude Code, Cursor, etc.

First Composition

Create index.html in your project root:

<div id="root"
     data-composition-id="hello"
     data-width="1920"
     data-height="1080">

  <h1 id="title" class="clip"
      data-start="0"
      data-duration="5"
      data-track-index="0"
      style="font-size: 96px; color: white; font-family: sans-serif;">
    HyperFrames
  </h1>
</div>
npx hyperframes preview   # see the result live
npx hyperframes render     # save as output.mp4

That's it. No bundler. No npm install of framework core. A single HTML file is a complete video composition.


Advanced Usage

GSAP Timeline Animations

GSAP animations work in HyperFrames with one critical rule — pause them so HyperFrames controls playback:

// timeline.js — loaded via <script> in your composition
const tl = gsap.timeline({ paused: true });
tl.to('#title', { x: 200, opacity: 1, duration: 2, ease: 'power2.out' })
  .to('#subtitle', { y: 0, opacity: 1, duration: 1 }, '-=0.5');

// Register so HyperFrames can seek it
window.__timelines = { main: tl };

HyperFrames seeks window.__timelines.main to precise frame positions during render. You can scrub through the timeline in browser preview at any moment and see exact frame-accurate preview.

Multi-Track Compositions

Track indices control layering in the compositor. A higher data-track-index renders on top:

<!-- Track 0: base video -->
<video data-track-index="0" ... />

<!-- Track 1: overlay graphics -->
<svg data-track-index="1" ... />

<!-- Track 2: captions (top of stack) -->
<div class="clip" data-track-index="2" ... />

Up to all available browser compositor layers per frame; no hard per-project limit.

Website-to-Video Pipeline

HyperFrames includes a website-to-hyperframes skill that captures a live URL and converts it into a video composition. This enables fully automated workflows like:

  • "Render the homepage at aratech.ae as a 30-second scrolling screencast video"
  • "Take the Notion doc and turn every section into a timed slide"
  • "Generate a product demo from the live staging URL"

This is the CI-friendly video generation loop that most marketing teams still do by hand.

Lottie and Three.js

// Lottie player registered by HyperFrames runtime
window.__hfLottie.play('animation.json');

// Three.js scene — HyperFrames injects current frame time
window.__hfThreeTime = 0; // seconds, set each frame by the engine

Adapter skills (/lottie, /three) auto-initialize the correct global and expose frame timing, so you don't need to set up any plumbing yourself.


Comparison: HyperFrames vs Remotion

HyperFrames is directly inspired by Remotion — the React-based programmatic video framework used at HeyGen in production. Both drive headless Chrome. Both produce deterministic frame-accurate output. Both ship agent skills. The architectural split is on one core decision: what does the primary author write?

HyperFramesRemotion
Primary authoring formatHTML + CSS + JSReact TSX
Build stepNone — index.html plays as-isRequired bundler
Library animations (GSAP, Lottie)Seekable, frame-accurate via adaptersPlays at wall-clock during render
Existing HTML/CSSPaste and animateMust be rewritten as JSX
Distributed renderingSingle machine todayLambda, production-ready
LicenseApache 2.0 (fully open source)Source-available, paid above small-team

For teams that already write HTML/CSS or need to embed existing web assets into video: HyperFrames is the obvious choice. For teams already deeply invested in a React component library with CI/CD: Remotion has some advantages on the distributed front.

The licensing gap is worth emphasizing. Remotion's "source-available" model means your company size, revenue, or deployment topology may trigger a paid license. HyperFrames has no such restriction under Apache 2.0.


Why This Matters Now

Two converging shifts make HyperFrames relevant this year rather than some future pipeline:

1. AI agent video is here. Claude Code, Cursor, and Codex can code. Now, with HyperFrames, they can code video. Natural language → HTML composition → rendered MP4 is an end-to-end workflow that runs today. You don't need to wait for an AI video model to be good enough; you need HTML and a renderer to be good enough — and they are.

2. Programmatic video is a competitive moat. Teams that generate video at scale from structured data (CSV → bar chart races, PDF → pitch videos, website → demo reels) have an inhuman advantage in content production velocity. The tools that lock in the most developers and agents first will define the infrastructure for the next wave of automated content.

HyperFrames is early but moving fast — 145 releases, a dedicated skills ecosystem, and integration with every major AI coding agent. It's the right time to start building with it.


Conclusion & Next Steps

HyperFrames isn't just another video tool. It's the first rendering framework that treats video composition as a first-class artifact in a developer's workflow — versioned in Git, reviewed in PRs, iterated by agents, and rendered deterministically on every commit.

If you've ever needed to:

  • Generate 100 personalized intro videos from CRM data
  • Turn a set of landing pages into a demo reel
  • Automate social clip creation from podcast transcripts
  • Embed AI-generated video output in a CI/CD pipeline

…HyperFrames is the tool that closes the loop.

Ready to get started?


Related Articles

  • Vibe-Coding a Backend in 2026: How Directus and the nikola66 Skill Turn Natural Language Into Production Schemas
  • OpenClaw vs Hermes Agent: Which AI Agent Framework Wins?

Table of Contents

  • ↗Table of Contents
  • ↗Introduction
  • ↗The Problem With Programmatic Video Today
  • ↗The Solution: HyperFrames
  • ↗Key Features
  • ↗HTML-Native Composition
  • ↗AI-First Design
  • ↗50+ Ready-to-Use Blocks
  • ↗The Frame Adapter Pattern
  • ↗Deterministic Rendering Engine
  • ↗Fully Open Source (Apache 2.0)
  • ↗Getting Started
  • ↗Prerequisites
  • ↗Quick Setup
  • ↗Option 2: AI agent (recommended)
  • ↗Then use /hyperframes in Claude Code, Cursor, etc.
  • ↗First Composition
  • ↗Advanced Usage
  • ↗GSAP Timeline Animations
  • ↗Multi-Track Compositions
  • ↗Website-to-Video Pipeline
  • ↗Lottie and Three.js
  • ↗Comparison: HyperFrames vs Remotion
  • ↗Why This Matters Now
  • ↗Conclusion & Next Steps
  • ↗Related Articles

Related Posts

8 Open-Source AI Tools You Missed This Week

From context compression to cross-platform AI agents — here are 8 powerful open-source projects reshaping the AI development landscape.

Necolas HamwiNecolas Hamwi
June 12, 2026 - 10 min read
OpenAI Dreaming V3 concept art - ChatGPT autonomous memory architecture

OpenAI's 'Dreaming V3' — ChatGPT Finally Has Persistent Memory

OpenAI's Dreaming V3 replaces saved memories with an autonomous background synthesis system. Factual recall jumps from 41.5% to 82.8%, compute drops 5x, and ChatGPT finally remembers like a thinking partner — not a notepad.

Necolas HamwiNecolas Hamwi
June 10, 2026 - 8 min read
Claude Fable 5: Anthropic Brings Mythos-Class Intelligence to the Public

Claude Fable 5: Anthropic Brings Mythos-Class Intelligence to the Public

Anthropic launched Claude Fable 5, the first publicly available Mythos-class model, delivering state-of-the-art intelligence with safety guardrails.

Necolas HamwiNecolas Hamwi
June 9, 2026 - 11 min read